This DIP outlines some changes to clean up the exception catching syntax in the D language
According to the grammar documentation on dlang.org, the following code:
catch (A) .someFunction();
should be parsed with the LastCatch rule, with (A).someFunction(); being treated as a function call.
What actually happens, is that DMD parses this using the Catch rule, treating A as an exception type and then calling the global-scoped someFunction().
This is caused by several bugs:
- Exceptions are not required to have names by the compiler. The grammar specification requires that exceptions be given names, although the specification also states "If just type T is given and no variable v, then the catch clause is still executed.". In this regard the specification is not consistent with itself.
- The Catch rule is assumed whenever a "(" token is hit after a "catch" token, regardless of whether it starts a NoScopeNonEmptyStatement or not.
- The CatchParameter rule is rewritten as follows:
CatchParameter: BasicType Identifier?This change merely updates the documentation to be consistent with DMD's treatment of exception names.
- The LastCatch rule is removed. Users who want to keep the functionality of their code unchanged must now explicitly "catch (Throwable)". This change is necessary to prevent the grammar from becoming ambiguous.
Code Breakage and Upgrade Path
Code that used the LastCatch syntax will no longer compile.
A tool that can automatically convert usage of this syntax to the more common "catch (Throwable)" syntax is located here: https://gist.github.com/Hackerpilot/5ff6d86f4d22a14a00f3
This document has been placed in the Public Domain.